CPS: Synergy: Learning to Walk - Optimal Gait Synthesis and Online Learning for Terrain-Aware Legged Locomotion

点击翻译按钮获取中文摘要

Legged robots have captured the imagination of society at large, throughentertainment and through the dissemination of research findings. Yet,today's reality of what (bipedal) legged robots can do falls short ofsociety's vision. A big part of the reason is that legged robots areviewed as surrogates for humans, able to go wherever humans can as aidsor as assistants where it might also be too dangerous or risky. It isin the expectation of robustness and walking facility that today'sresearch hits its limits, especially when the terrain has granularproperties. Impeding progress is the lack of a holistic approach to thecyber-physical modeling and control of legged robots. The vision ofthis work is to unite experts in granular mechanics, optimal control,and learning theory in order to define a methodology for advancingcyber-physical systems (CPS) involving a tight coupling of the physical withthe cyber through dynamic interactions that must be learned online. Theproposed work will advance the science of cyber-physical systems by moreexplicitly tying sensing, perception, and computing to the optimizationand control of physical systems whose properties are variable anduncertain. Achieving reliable, adaptive legged locomotion over terrainwith arbitrary granular properties would transform several applicationdomain areas of robotics; e.g., disaster response, agricultural andindustrial robotics, and planetary robotics. More broadly, the sametools would apply to related CPS with regards to terrain awareexoskeleton and rehabilitation prosthetics for persons with missing,non-functional, or injured legs, as well as to energy networks withtime-varying, nonlinear dynamics models.The CPS platform to be studied is that of a bipedal robot locomotingover granular ground material with uncertain physical properties (sand,gravel, dirt, etc.). The proposed work seeks to overcome currentimpediments to reliable legged locomotion over uncertain terrain type,which fundamentally relies on the controlled interaction of the robot'sfeet with the physical environment. The research goal is to improve theperception and control of legged locomotion over granular media for theexpress purpose of achieving robust, adaptive, terrain-aware locomotion.It revolves around the hypothesis that simple models with decentpredictive performance and low computational overhead are sufficient forthe optimal control formulations as the compute-constrained adaptivesubsystem will both learn and classify the peculiarities of the terrainonline. The main research objectives will involve: [1] a validatedco-simulation platform for legged robot movement over granular media;[2] terrain-dependent, stable gait generation and gait transitionstrategies via optimal control; [3] online, compute-constrained learningof granular interactions for adaptation and terrain classification; and[4] validated contributions using experimental testbeds involvingvariable and unknown (to the robot) granular media. Given the highvalue of the robotic platforms and the research with regards to outreachand participation, they will be used as outreach tools and to create neweducational modules for promotion of STEM fields. Further, themulti-disciplinary nature of the work will be highlighted in order toemphasize its importance.